Hello WKND AI Warriors!

The US Department of War blacklists Anthropic for its refusal to compromise on Claude being used for domestic mass surveillance and autonomous weapon systems.

Also, Meta’s AI alignment director let the viral OpenClaw agent go rogue and delete her emails.

Plus, a blog about Anthropic’s Claude being able to code COBOL sent IBM’s stock plummeting, losing $30B in one day.

Oh yeah, and while Jack Dorsey slashes 4,000 jobs because of AI, Burger King’s AI “Patty” monitors how many times employees say “please” and “thank you”.

Today’s newsletter includes:

• 🎓 AI COURSE OF THE WEEK

• 📰 AI NEWS RECAP

• 🤿 AI DEEP DIVE

• 📝 AI PROMPT OF THE WEEK

• 🎨 AI IMAGE OF THE WEEK

🎓 AI COURSE OF THE WEEK

👉 Get instant access here.

📰 AI NEWS RECAP

Lorem

Anthropic just lost access to US government contracts.

Not because Claude underperformed.
Because the company refused to remove two restrictions.

According to reports, the Pentagon wanted broad discretion to use Claude for any lawful purpose.

Anthropic asked for written limits:

No domestic mass surveillance.
No fully autonomous weapons.

There was no agreement.

Talks stalled.
The company was labeled a supply chain risk.
Federal ties were cut.

That is the sequence.

The interesting part is not the politics.
It is the signal.

For years, AI labs talked about red lines.
Now one of them tested those lines in a real procurement negotiation.

A private company said no.

The government responded with exclusion.

You can read this in different ways.

As principle.
As strategy.
As leverage.

What matters is that the tension is no longer theoretical.

Frontier AI companies are no longer just technology vendors.
They sit in the middle of national security, public policy, and commercial markets.

When ethics meet contracts, something has to give.

What will it be?

Time will tell.

The same day Anthropic was cut off, OpenAI signed a deal to deploy its models on the Department of War's classified network. Altman agreed to let the Pentagon use OpenAI for any lawful purpose while promising safeguards against domestic surveillance and autonomous weapons.

Anthropic published a blog showing Claude Code can read, write, and modernize legacy COBOL systems, directly threatening IBM's mainframe services business with banks, airlines, and governments. IBM's stock dropped 13% in a single day, wiping roughly $30 billion in market cap, its worst single-day drop in 26 years.

Meta's director of AI alignment, Summer Yue, connected the open-source agent OpenClaw to her real work inbox, only to watch it announce it would "trash EVERYTHING older than Feb 15." She tried typing "STOP OPENCLAW" from her phone but couldn't halt it in time, forcing a sprint to her computer to pull the plug, prompting her to admit it was a "rookie mistake" and that "alignment researchers aren't immune to misalignment."

Jack Dorsey announced Block is cutting roughly 4,000 of its 10,000 employees, explicitly blaming AI rather than a business downturn. Block's gross profit is actually up 24% and its stock jumped 20% on the news, making it one of the clearest signals yet that profitable companies will cut headcount as AI absorbs the work.

Anthropic published research showing that Chinese frontier models have been systematically querying Claude at scale to clone its capabilities into cheaper, unaligned copies. The paper outlines new detection and countermeasures designed to block these distillation attacks before the stolen knowledge can be used without Claude's safety training.

Burger King's new "Patty" AI listens to employees throughout their shift, grading their courtesy, flagging broken equipment, and monitoring stock levels in real time. For those keeping a dystopia checklist, the corporation now has an always-on AI scoring whether workers say "please" enough.

🤿 AI DEEP DIVE

What are the implications of the US government blacklisting Anthropic?

Watch it here!

How can you help?

Refer my newsletter to help others learn AI.

Missed last week’s edition?

Find all of my newsletters here!